Synaptic Plasticity and Addiction 2007

REVIEWS

Synaptic plasticity and addiction

Julie A. Kauer* and Robert C. Malenka‡ Abstract | Addiction is caused, in part, by powerful and long-lasting memories of the drug experience. Relapse caused by exposure to cues associated with the drug experience is a major clinical problem that contributes to the persistence of addiction. Here we present the accumulated evidence that drugs of abuse can hijack synaptic plasticity mechanisms in key brain circuits, most importantly in the mesolimbic dopamine system, which is central to reward processing in the brain. Reversing or preventing these drug-induced synaptic modifications may prove beneficial in the treatment of one of society’s most intractable health problems.

Long-term potentiation More than a century ago, Ramon y Cajal speculated that Addiction is not triggered instantaneously upon (LTP). Activity-dependent information storage in the brain results from alterations exposure to drugs of abuse. It involves multiple, com- strengthening of synaptic in synaptic connections between neurons1. The discov- plex neural adaptations that develop with different transmission that lasts at least ery in 1973 of long-term potentiation (LTP) of glutamate time courses ranging from hours to days to months one hour. synapses in the hippocampus2 launched an exciting (BOX 1). Work to date suggests an essential role for Long-term depression exploration into the molecular basis and behavioural synaptic plasticity in the VTA in the early behavioural (LTD). Activity-dependent correlates of synaptic plasticity. Partly because LTP was responses following initial drug exposures, as well as in weakening of synaptic first described at synapses in the hippocampus, a brain triggering long-term adaptations in regions innervated transmission that lasts at least 9 one hour. region necessary for declarative memory formation, by dopamine (DA) neurons of the VTA . By contrast, there was an early assumption that synaptic plasticity downstream synaptic changes in the NAc and other represents a cellular building block used exclusively for brain regions, are likely to represent the formation of learning and memory. However, it has since become powerful and persistent links between the reinforcing clear that LTP and its counterpart, long-term depression aspects of the drug experience and the multiple cues (LTD), are basic properties of most excitatory syn- (both internal and external) associated with that exp- apses throughout the CNS, and are used for multiple erience5–10. Here we review emerging evidence that brain functions in addition to learning and memory3. addictive drugs elicit or modify synaptic plasticity in For example, LTP and LTD appear to be essential in many of the key brain regions involved in addiction, the stabilization and elimination of synapses during the and that these synaptic modifications have important developmental fine-tuning of neural circuits in many behavioural consequences. A major motivation for areas of primary sensory cortex4. this research is the assumption that addictions to the It therefore may not be surprising that evidence accu- different classes of abused substances share important mulated over the last decade demonstrates that drugs of underlying brain mechanisms. Identifying these mecha- abuse can co-opt synaptic plasticity mechanisms in brain nisms will advance our ability to treat and prevent these circuits involved in reinforcement and reward process- often devastating disorders, as well as other related *Department of Molecular Pharmacology, Physiology ing. Indeed, an influential hypothesis is that addiction behaviours, such as gambling. and Biotechnology, represents a pathological, yet powerful, form of learning Of course, the brain adaptations that underlie addic- Brown University, and memory5–10. Although the brain circuitry underlying tion are complex and involve drug-induced changes Providence, Rhode Island addiction is complex, it is unequivocal that the mesolim- in essentially every parameter that has been studied 02912, USA. ‡Nancy Pritzker Laboratory, bic dopamine system, consisting of the ventral tegmental including gene transcription, membrane excitability and Department of Psychiatry area (VTA) and nucleus accumbens (NAc), as well as neuronal morphology. Moreover, because of advances and Behavioural Sciences, associated limbic structures (FIG. 1), are critical substrates in our understanding, and the societal importance, of Stanford University School for the neural adaptations that underlie addiction. It is the neurobiology of addiction, this topic has been the of Medicine, Stanford, also clear that the interactions between addictive drugs subject of numerous reviews in both the basic science California 94304, USA Correspondence to R.C.M. and synaptic plasticity in different brain regions will con- and clinical literatures. Thus, we will intentionally limit e-mail: [email protected] tribute to specific aspects of addiction, such as craving, our discussion primarily to those studies that most doi:10.1038/nrn2234 withdrawal and, perhaps most importantly, relapse. directly demonstrate drug-induced modulation of

Conditioned place synaptic plasticity mechanisms and try to construct a repeated exposure. Given the critical role that NMDAR- preference coherent picture from an often confusing and, as yet, dependent synaptic plasticity is thought to have in nor- A behavioural task during incomplete literature. mal learning and memory13, these findings immediately which a subject learns to suggested that processes akin to associative learning are associate the drug experience Addiction and learning with a specific physical essential in the early development of addiction. 25–27 environment. A subject will Synaptic plasticity is required for neuroadaptations that Abundant additional evidence supports the choose to spend more time in result from a wide range of environmental stimuli. It was notion that excitatory synaptic function within meso- an environment in which it therefore attractive to hypothesize that drugs of abuse limbic dopamine circuits is crucial for the behavioural previously had a ‘rewarding’ cause long-term changes on behaviour by altering syn- responses to drugs of abuse. Furthermore, human experience and less time in an environment in which it had aptic function and plasticity in relevant brain circuits. imaging studies in addicted subjects demonstrate the an aversive experience. Moreover, data from diverse behavioural experiments powerful cognitive and emotional effects of cues that with drugs of abuse has implicated specific signalling were previously associated with the drug experience28. molecules already identified as key players in LTP and Preventing relapse is the major clinical problem in the LTD at other synapses10. Indeed, accumulating evi- treatment of addiction, suggesting the need to under- dence links various behavioural models of key features stand the cellular nature of the powerful ‘memories’ of addiction with synaptic plasticity in brain areas that caused by prior drug experiences. Thus, experimental process reinforcement and reward. work in animal models, as well as clinical studies, pro- Studies demonstrating that blocking !-methyl-"- vides compelling support for the importance of learning aspartate receptors (NMDARs) could short-circuit the and memory mechanisms in addiction. development of drug-induced behavioural adaptations in certain addiction models were among the first clues LTP and LTD mechanisms that addictive drugs might access the same processes that A ubiquitous property of all synapses is their ability are used to store learned information. For example, to undergo activity-dependent changes in synaptic NMDAR blockade, known to prevent many forms of strength, that is, synaptic plasticity. Much of the LTP and LTD in other brain regions3, also prevents mechanistic work on long-term synaptic plasticity in conditioned place preference, behavioural sensitization the mammalian brain over the last few decades has and self-administration of drugs of abuse14–21 (BOX 2). focused on the forms of LTP and LTD observed at excita- Furthermore, NMDAR blockade specifically within the tory synapses, although it is now clear that inhibitory VTA (but not the NAc) effectively prevents both behav- synapses can exhibit LTP and LTD as well. Synaptic ioural sensitization and conditioned place preference, plasticity can be studied most effectively using electro- supporting the idea that NMDAR-dependent processes physiological methods in brain slices that are viable for in the VTA might have a pivotal role in the development several hours, and therefore, the cellular mechanisms of addiction15–17. Importantly, NMDAR blockade does underlying the first few hours of LTP and LTD are the not prevent the acute locomotor response to psycho- best understood. Before discussing the interactions stimulant drugs, only the sensitization that occurs with between drugs of abuse and long-term synaptic plasticity, it is useful to review our mechanistic understanding of the most common forms of LTP and LTD (FIG. 2). Only by understanding these core synaptic mechanisms PFC can we hope to understand how drugs of abuse usurp Hippocampus or modify them.

NMDAR-dependent LTP. NMDAR-dependent LTP, Striatum first observed in the hippocampus, has been intensively LDTg examined for over three decades and remains the best understood form of long-lasting synaptic plasticity in the mammalian brain2,3 (FIG. 2a). It requires the NAc VP VTA activation of NMDARs by presynaptically released BNST LH glutamate when the postsynaptic membrane is significantly depolarized. This relieves the voltage-dependent AMG block of the NMDAR by Mg2+, allowing Ca2+ to enter Figure 1 | Mesolimbic dopamine system circuitry. Simplified schematic of the circuitry postsynaptic dendritic spines. The rise in postsynaptic of the mesolimbic dopamine system in the rat brain highlighting the major inputs to the Ca2+ concentration, the crucial trigger for LTP, activates nucleus accumbens (NAc) and ventral tegmental area (VTA) (glutamatergic projections, complex intracellular signalling cascades that include blue; dopaminergic projections, red; GABAergic projections,Na turorange;e Revie orexinergicws | Neuroscienc e several protein kinases, most notably CaMKII29. The projections, green). Glutamatergic synapses excite postsynaptic neurons and GABAergic primary mechanism underlying the increase in synaptic synapses inhibit postsynaptic neurons. Dopamine release exerts more complex strength during LTP is a change in α-amino-3-hydroxy- modulatory effects. The release of dopamine from VTA neurons increases in response to administration of all drugs of abuse5–10,50. These neurons also fire in response to novelty 5-methyl-4-isoxazole propionic acid receptor (AMPAR) and their firing patterns may encode a prediction signalling the reward value of a trafficking that results in an increased number of stimulus relative to its expected value143. AMG, amygdala; BNST, bed nucleus of the AMPARs in the postsynaptic plasma membrane with stria terminalis; LDTg, laterodorsal tegmental nucleus; LH, lateral hypothalamus; no effect on NMDARs3,29. Within a few hours, the main- PFC, prefrontal cortex; VP, ventral pallidum. tenance of LTP requires protein synthesis30, and there is

Induction of synaptic growing evidence that LTP is accompanied by observ- for LTP, including serine/threonine phosphatases, which plasticity able enlargements of dendritic spines and associated dephosphorylate critical synaptic substrates, including Refers to the cellular postsynaptic densities3,31,32. These structural changes the AMPARs themselves3 (FIG. 2c). The depression of mechanisms required for the may be essential to cement the information-storage synaptic strength during NMDAR-dependent LTD is due events initiating or triggering LTP or LTD. process initiated at synapses upon LTP induction. It to the removal of synaptic AMPARs via dynamin- and should also be noted that an alteration in the trafficking clathrin-dependent endocytosis3,39. An intriguing feature or numbers of NMDARs at synapses could potentially of NMDAR-dependent LTD is that NMDAR-mediated alter the threshold for induction of NMDAR-dependent synaptic responses are also depressed by mechanisms LTP (and LTD; see below). A subunit switch between that are distinct from those responsible for the LTD particular NMDAR subunits can also up- or downregu- of AMPAR-mediated responses40,41. This observation late NMDAR-mediated synaptic currents so that more suggests that after this form of LTD is induced, further or less Ca2+ will enter the postsynaptic neuron during NMDAR-dependent synaptic plasticity will be limited, receptor activation, thereby altering the induction of at least temporarily. synaptic plasticity33. Metabotropic glutamate receptor-dependent LTD. Presynaptic LTP. A distinct form of LTP was first Activation of metabotropic glutamate receptors (mGluRs) described at synapses between the mossy fibres of the can also lead to a postsynaptically induced and expressed dentate granule cells and area CA3 hippocampal pyra- LTD; this was first described at parallel fibre synapses midal neurons, but similar examples have been found in on cerebellar Purkinje cells42 (FIG. 2d). Other forms of the neocortex and cerebellum3,34. This form of LTP does mGluR-dependent LTD using somewhat overlapping not require NMDARs and postsynaptic factors may not cellular mechanisms have subsequently been described be required (although this remains controversial35,36). in the hippocampus and the neocortex3. At the parallel Instead, presynaptic LTP appears to be initiated by an fibre synapse, LTD is associative, requiring both post- activity-dependent rise in intracellular Ca2+ within the synaptic Ca2+ influx through voltage-gated ion channels presynaptic terminals (FIG. 2b). The Ca2+ rise activates and postsynaptic group I mGluR activation, whereas at adenyl cyclases to produce cyclic AMP, with subsequent other synapses, activation of postsynaptic mGluRs alone activation of protein kinase A (PKA)3,34. This in turn appears to be sufficient. In most cases, however, this form leads to a persistent increase in the amount of glutamate of LTD is mediated by clathrin-dependent endocytosis of released each time an action potential reaches the nerve synaptic AMPARs. Interestingly, at certain developmen- terminal. Rab3A and RIM1A, proteins that act to coor- tal stages, rapid protein synthesis is required for both dinate synaptic vesicle interactions with the presynaptic mGluR-triggered AMPAR endocytosis and LTD43. active zone, have an essential role in the increased glutamate release37,38. Endocannabinoid-mediated LTD. At many CNS glutamatergic and G-aminobutyric acid (GABA)-releasing NMDAR-dependent LTD. NMDAR-dependent LTD synapses, a brief, strong postsynaptic Ca2+ influx (and in is induced by weak activation of NMDARs (for exam- some cases activation of mGluRs or muscarinic receptors ple, due to modest membrane depolarization or low alone) triggers the synthesis of endocannabinoids (eCBs), stimulation frequencies) and is thought to result from lipophilic molecules that travel retrogradely across the a smaller rise in postsynaptic Ca2+ than is required for synapse to bind to presynaptic CB1 receptors and trans- LTP3. This triggers a different subset of Ca2+-dependent iently depress neurotransmitter release for a period of intracellular signalling molecules than those required many seconds44 (FIG. 2e). At some synapses, however,

Box 1 | Drug administration protocols There are many different ways to administer drugs of abuse. A drug may be administered by the investigator (passive administration) or an animal can be trained to self-administer the drug in response to cues. The time course over which drugs are administered and the time point at which assays are performed, during drug administration or following drug withdrawal, are also parameters that are under experimental control. These details are important because neuroadaptations to drugs of abuse occur over varying timescales and can be greatly influenced by the mode and duration of drug administration. An additional complexity is that distinct behavioural and neurobiological results can ensue depending on the novelty or other features of the environment in which the animal receives the drug. For example, behavioural sensitization is much more robust when the investigator-administered drug is given in a novel cage compared with the home cage11. Recent studies also have found differences between groups of self-administering animals depending on the length of time the drug is available each day12. By varying drug administration protocols, investigators may selectively highlight particular aspects of the drug experience and the processes contributing to addiction. Assays at relatively early time points can detect changes related to tolerance or the symptoms of acute withdrawal, as well as changes underlying the development of craving. Adaptations related to craving and relapse may be most apparent when time points of days or weeks after drug withdrawal are examined. In some studies, after a period of withdrawal, a subsequent drug dose is given (for example, for the assessment of a behavioural response) and this re-exposure to the drug itself can have significant effects on synaptic and circuit properties. It is important to note carefully the administration protocol and the assay time points as contradictory results can be obtained from minor differences in these variables.

Box 2 | Behavioural sensitization example, excitatory inputs from the PFC selectively form synapses onto VTA dopaminergic neurons that project Sensitization is the gradually escalating behavioural and motivational response to a back to the PFC but not onto neighbouring dopaminergic fixed drug dose. Two features of sensitization are intriguing in the context of drug cells in the VTA that project to the NAc52. addiction: multiple different addictive drugs produce sensitization, and after cessation Before addressing the question of whether drugs of of drug exposure, sensitization routinely lasts for weeks or months. Behavioural sensitization is most commonly assayed as drug-induced increases in locomotor abuse can trigger synaptic plasticity in the VTA, it was activity22 and this has been associated with enhancement of the rewarding properties of important to establish that phenomena such as LTP drugs of abuse23. Results from the behavioural sensitization model were some of the first and LTD do in fact occur at VTA synapses. Indeed, to hint that NMDAR-dependent synaptic plasticity might be an essential contributor to excitatory synapses on VTA DA cells express a form the neural adaptations leading to drug addiction24,25. There is strong evidence that the of NMDAR-dependent LTP53–56, as well as LTD that, ventral tegmental area is involved in the triggering of behavioural sensitization, whereas surprisingly, requires voltage-dependent Ca2+ channels, the nucleus accumbens is crucial for its expression, but not its triggering. not NMDARs57,58. Furthermore, an mGluR-dependent LTD has also been reported at VTA synapses59, and is described in the next section. These findings set the stage prolonged eCB release instead causes LTD, which is for a study that directly tested whether in vivo adminis- mediated by a long-lasting depression of transmitter tration of an addictive drug produced long-term changes release (eCB-LTD)45. Why eCB release produces only a at excitatory synapses on VTA DA neurons60. To monitor transient synaptic depression at some synapses while at changes in excitatory synaptic strength, the investiga- other synapses persistent LTD is elicited is not fully under- tors measured the ratio of AMPAR-mediated excitatory stood. Recent work suggests that the presynaptic mecha- postsynaptic currents (EPSCs) to NMDAR-mediated nisms underlying the transient depression due to eCBs EPSCs (the AMPAR/NMDAR ratio) (FIG. 3), and found versus eCB-LTD differ, with PKA and RIM1A dependent that a single exposure to cocaine caused a large increase signalling being necessary only for eCB-LTD46. in this ratio in VTA DA cells when measured 24 hours later in brain slices. Additional assays indicated that this Homeostatic synaptic scaling. In addition to LTP and cocaine-induced change, like NMDAR-dependent LTP, LTD, which usually are synapse specific, synaptic strength was due to an upregulation of AMPARs and potentially can be modified when activity levels are changed for pro- shared mechanisms with the synaptically evoked LTP longed periods (hours to days). Specifically, prolonged elicited in VTA slices. Furthermore, this drug-induced decreases in activity globally increase synaptic strength LTP was prevented when animals were pre-treated with whereas prolonged increases in activity decrease synaptic an NMDAR antagonist. strength47. These widespread changes in synaptic strength These findings support the hypothesis that in vivo are thought to be homeostatic responses that maintain cocaine exposure elicits LTP at excitatory synapses on the activity of individual cells within some finite range VTA DA neurons. An obvious yet critical question is while keeping constant the relative differences in strength whether other drugs of abuse cause the same synaptic between synapses, caused by LTP and LTD, constant. modification. The finding that application of nicotine Most evidence suggests that synaptic scaling is caused could evoke LTP at VTA DA excitatory synapses55 is by changes in synaptic AMPAR content together with consitant with this idea. Further experiments show a presynaptic changes in transmitter release48. Mechanisms similar potentiation of the AMPAR/NMDAR ratio 24 for this form of synaptic plasticity are currently under hours after in vivo administration of a number of diverse investigation and might include changes in local protein addictive drugs, including amphetamine, morphine, synthesis, nerve growth factors or diffusible factors, such ethanol and nicotine61 (FIG. 4). The increased ratio could as tumour necrosis factor A (TNFA)49. be detected within two hours of amphetamine exposure in vivo, as expected if LTP is the underlying mecha- Drug exposure triggers LTP in the VTA nism62. Importantly, administration of widely used non- Different classes of drugs of abuse all increase the release addictive drugs (fluoxetine and carbamazepine), did of DA in the NAc50 and this convergence, along with not change the AMPAR/NMDAR ratio61. Furthermore, compelling evidence from behavioural studies, indicates no change in the AMPAR/NMDAR ratio was observed at that the mesolimbic DA system is required for drug hippocampal synapses or at excitatory synapses on VTA addiction5–10,22,23,28. The major cell type in the VTA is the GABAergic cells, indicating that the effect of cocaine at dopaminergic neuron, which receives excitatory inputs VTA DA cell synapses was specific60. The finding of an from the prefrontal cortex (PFC), laterodorsal tegmental increased AMPAR/NMDAR ratio after administra- nucleus and lateral hypothalamus51. Dopaminergic tion of multiple different classes of drugs of abuse with neurons are inhibited by local interneurons, which distinct molecular targets and differing behavioural Excitatory postsynaptic currents generate GABAA receptor-mediated responses, as well profiles suggests that this synaptic adaptation (that is, (EPSCs). Currents measured as by GABAergic projections from the NAc and ventral LTP at excitatory synapses on VTA DA neurons) might using electrophysiological pallidum. VTA DA neurons themselves provide major be directly related to the addictive properties of these recordings from a single neuron projections to the NAc and PFC. As many as 35% of VTA compounds. while electrically stimulating neurons are GABAergic, and in addition to providing Stress is a potent trigger of relapse in humans and axons to release 63–66 neurotransmitter. For the local inhibition, these neurons also project to the NAc many animal addiction models . This observation purposes of this Review, EPSCs and PFC (FIG.1). Precise anatomical relationships exist provided the motivation to test whether acute stress also are glutamate-mediated. between neurons in the VTA and projection targets. For increased the AMPAR/NMDAR ratio in DA neurons.

a NMDAR-dependent LTP b Presynaptic LTP c NMDAR-dependent LTD Presynaptic 2+ terminal Ca Voltage-gated Ca2+ channels cAMP PKA Glu Rab3a RIM1 AC

Mg2+ Mg2+ NMDAR AMPAR NMDAR AMPAR NMDAR AMPAR

Ca2+ Ca2+ Dendritic spine CaMKII Calcineurin PP1 Postsynaptic dendrite Expression: postsynaptic Expression: increased presynaptic Expression: internalization of insertion of AMPARs neurotransmitter release postsynaptic AMPARs

d mGluR-dependent LTD e eCB-LTD

CB1R

eCB NMDAR AMPAR NMDAR AMPAR

mGluR mGluR 1/5 eCB PLC 1/5 Voltage-gated ? Ca2+ ? Ca2+ channels Ribosome

Expression: internalization of Expression: decreased presynaptic postsynaptic AMPARs neurotransmitter release Figure 2 | Well-described forms of LTP and LTD. Highly simplified diagrams of the induction and expression of synaptic plasticity observed in the rodent brain. a | !-methyl-"-aspartate receptor (NMDAR)-dependentNa turlong-terme Reviews potentiation | Neuroscienc e (LTP) has been observed in many different brain regions and is dependent on postsynaptic NMDAR activation and calcium/calmodulin-dependent protein-kinase II (CaMKII) for its initiation3. The voltage-dependent relief of the magnesium block of the NMDAR channel allows the synapse to detect coincident presynaptic release of glutamate (Glu) and postsynaptic depolarization. A-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptor (AMPAR) insertion into the postsynaptic membrane is a major mechanism underlying LTP expression. b | Presynaptic LTP has been best characterized at mossy fibre–CA3 hippocampal synapses as well as at parallel fibre–Purkinje cell cerebellar synapses3,34. Repetitive synaptic activity leads to the entry of presynaptic Ca2+, which activates a Ca2+-sensitive adenylate cyclase (AC) leading to a rise in cAMP and the activation of cyclic AMP-dependent protein kinase A (PKA). This in turn modifies the functions of Rab3a and RIM1A leading to a long-lasting increase in glutamate release34,37,38. Involvement of postsynaptic signalling molecules (not shown) has also been reported35,36. c | NMDAR-dependent long-term depression (LTD) is triggered by Ca2+ entry through postsynaptic NMDAR channels, leading to increases in the activity of the protein phosphatases calcineurin and protein phosphatase 1 (PP1). The primary expression mechanism involves internalization of postsynaptic AMPARs and a downregulation of NMDARs by an unknown mechanism3,41. d | Metabotropic glutamate receptor (mGluR)-dependent LTD has been best characterized at cerebellar parallel fibre–purkinje cell synapses and Expression of synaptic hippocampal synapses. Activation of postsynaptic mGluR1/5 triggers the internalization of postsynaptic AMPARs, a plasticity process that under some conditions appears to require protein synthesis43. e | Endocannabinoid-LTD is the most Refers to the cellular recently discovered form of LTD, and has been observed in many brain regions. Either mGluR1/5 activation, leading to mechanisms responsible for 2+ maintaining a change in activation of phospholipase C (PLC) or an increase of intracellular Ca (or both), in the postsynaptic neuron initiates the synaptic strength, for example, synthesis of an endocannabinoid (eCB). The eCB is subsequently released from the postsynaptic neuron, travels an increase in neurotransmitter retrogradely to bind to presynaptic cannabinoid 1 receptors (CB1R) and this prolonged activation of CB1Rs depresses release. neurotransmitter release via unknown mechanisms45.

Indeed, 24 hours after a cold water swim (a manipu- The studies reviewed thus far provide strong evi- lation commonly used to elicit stress in rodents), the dence that drugs of abuse interact with synaptic plas- ratio was increased and, like the response to cocaine, ticity mechanisms in VTA DA neurons. In fact, in vivo the increase was blocked by a preceding dose of an administration of amphetamine not only elicits LTP at NMDAR antagonist61. This result raised the possibility excitatory VTA synapses, but also blocks LTD at the that administration of drugs of abuse elicited a stress same synapses when applied to slices, an effect that may response that was responsible for the observed drug- contribute to its potentiating effects57. Do these synaptic induced LTP. However, a glucocorticoid receptor antago- adaptations have any behaviourally relevant conse- nist, which blocked the potentiation caused by stress, quences? A standard way of beginning to address this did not block the potentiation by cocaine61. Conversely, issue is to examine the behaviour of mutant mice that a D1 dopamine receptor antagonist blocked the increase lack relevant forms of synaptic plasticity in the VTA and in the AMPAR/NMDAR ratio elicited by cocaine but not examine them for behavioural deficits. Current models of the increase caused by stress67. These results demonstrate NMDAR-dependent LTP support the idea that insertion that even though stress and cocaine elicit the same syn- of glutamate receptor 1(GluR1)-containing AMPARs is aptic adaptation in VTA DA neurons, they do so through an early necessary step in LTP expression68. Furthermore, distinct mechanisms. overexpression of GluR1 subunits in the VTA produced sensitized behavioural responses to morphine69. Thus, the electrophysiological and behavioural effects of cocaine a Basal/control state b LTP exposure in mutant mice lacking GluR1 were explored67. Administration of cocaine or an acute stress to GluR1 knock-out mice did not increase the AMPAR/NMDAR ratio measured 24 hours later, demonstrating a critical role for this AMPAR subunit in VTA LTP. Although the mutant animals still developed locomotor sensitization to cocaine (BOX 2), conditioned place preference to cocaine was absent, as was their conditioned increase in locomotor activity when placed in the activity box in which they had previously experienced cocaine67. These NMDAR AMPAR NMDAR AMPAR results are consistent with the idea that drug-induced LTP of excitatory synapses on VTA DA neurons might be necessary for attributing motivational significance to the drug experience or for the learned association between context and drug experience. Of course, in these studies GluR1 was absent throughout the brain and thus, more work will be needed to prove that the absence of LTP AMPAR/NMDAR ratio = 0.4 AMPAR/NMDAR ratio = 1.0 in the VTA, rather than other brain structures, caused the observed behavioural impairments. It is somewhat puzzling that GluR1 overexpression in the VTA produces behavioural sensitization on its own, suggesting +40 mV +40 mV that LTP-like changes in the VTA are sufficient to drive behavioural responses to morphine toward a sensitized AMPAR EPSC NMDAR EPSC phenotype69, whereas GluR1 knock-out animals exhibit normal behavioural sensitization to cocaine67. Further Figure 3 | Synaptic strength measured using the AMPAR/NMDAR ratio. The basal work will be necessary to determine whether these dif- strength of excitatory synapses is difficult to compare between different cells and ferences reflect important mechanistic differences in the preparations. Calculating the ratio of A-amino-3-hydroxy-5-methyl-4-isoxazoleNature Reviews | Neur oscience propionic acid receptor (AMPAR)-mediated synaptic currents to !-methyl-"-aspartate actions of cocaine and morphine, or rather are due to receptor (NMDAR)-mediated synaptic currents of a population of stimulated synapses is the different methodologies that were used to study the a normalization procedure that facilitates such comparisons because it is independent of role of GluR1. parameters such as the positioning of electrodes or the number of synapses that are What happens to excitatory synaptic function in VTA activated. It is commonly calculated by holding the membrane potential of cells at DA neurons after repeated exposure to cocaine — do the positive potentials (for example, +40 mV), to completely relieve the block of NMDARs by AMPAR/NMDAR ratios become even bigger, or is there magnesium, and measuring the amplitude of AMPAR excitatory postsynaptic currents a ceiling effect? Surprisingly, after seven daily cocaine (EPSCs) and NMDAR EPSCs. A typical procedure is to record a dual component EPSC injections, the AMPAR/NMDAR ratios remained at (mediated by both AMPARs and NMDARs; not shown) and then apply the NMDAR the same level seen 24 hours after a single injection70. antagonist D-APV to isolate the AMPAR EPSC. The NMDAR EPSC is obtained by digital subtraction of the AMPAR EPSC from the dual component EPSC. a | A synapse is The persistence of the potentiation was also similar in illustrated in the basal or control state. The isolated AMPAR and NMDAR components of both groups: ratios remained elevated five days after the EPSC are shown below the synapse diagram. b | After LTP induction, if AMPARs are the last cocaine injection but were near control levels inserted into the postsynaptic membrane, the AMPAR component of the EPSC is after ten days. Moreover, immediately following the first enhanced, whereas the NMDAR component remains unchanged, increasing the ratio. cocaine administration the precise AMPAR/NMDAR This method assumes that there are no significant changes in the proportion of rectifying ratio from a given animal correlated well with its drug- (Ca2+-permeable) AMPARs that would confound measurements made at +40 mV. induced locomotor behaviour, but after this time point

GABA Glutamate

Morphine

NMDAR AMPAR cGMP GC NO NOS + mGluR ? + Cocaine Chronic cocaine + Morphine decreases GABAergic + Amphetamines synapse function GABAAR + Nicotine + Ethanol + Stress

Figure 4| Drugs of abuse modulate synaptic function and plasticity in the ventral tegmental area (VTA). Diagram showing the major effects of drugs of abuse on synaptic plasticity in the VTA. Several classes ofNa drugsture Re ofvie abusews | Neur (indicatedoscienc e in the white box), as well as acute stress, elicit long-term potentiation (LTP) possibly by increasing postsynaptic A-amino-3- hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) at glutamatergic synapses60–62,67,70; metabotropic glutamate receptor-dependent long term depression (LTD) can reverse this LTP80,81. !-methyl-"-aspartate receptor (NMDAR) activation also activates nitric oxide synthase (NOS), which leads to the production of nitric oxide (NO). NO is a diffusible messenger that is released from postsynaptic neurons and activates guanylate cyclase (GC) in neighbouring presynaptic inhibitory terminals. A rise in cyclic GMP (cGMP) elicits a long-lasting increase in G-aminobutyric acid (GABA) 96 release at GABAA receptor-containing synapses (LTPGABA) . Morphine prevents LTPGABA by inhibiting the signalling of NO to guanylate cyclase96. Chronic cocaine decreases inhibitory synaptic transmission via unknown mechanisms56.

the locomotor response and AMPAR/NMDAR ratio drug-seeking behaviour, even after several weeks of became uncorrelated70. In a related study, AMPAR/ cocaine withdrawal73. Further work is required to test the NMDAR ratios were examined in mice that received idea that growth factors like BDNF may be regulatory both an acute stress and cocaine, and these ratios also molecules linking rapid changes at synapses caused did not increase above the level seen with either stimulus by drug exposure with longer-lasting modifications alone67. These results are consistent with the idea that of circuit activity. although potentiation of excitatory synapses on VTA DA The results summarized in this section provide strong neurons may initially contribute to the incentive value evidence that drugs of abuse or stress cause potentiation attributed to the drug or stress experience, adaptations in of excitatory synapses on VTA DA neurons. It is clear, downstream circuitry are likely to be more important for however, that this one synaptic adaptation alone does the longer-lasting behavioural changes associated with not predict that addiction will follow. Addiction rarely addiction (BOX 2). occurs after, for example, a single exposure to nicotine Another reported effect of chronic (5–7 day) cocaine or alcohol, yet one exposure to either drug potentiates administration is to enhance a form of LTP elicited by synapses on VTA DA neurons61. Furthermore, a single a spike-timing protocol56. However, we (P. Luu and acute stress does not lead to drug addiction despite the R. C. Malenka, unpublished observations) and others fact that this experience also potentiates VTA synapses61. (E. Argilli and A. Bonci, personal communication) have Instead, these experiments suggest that the LTP at excita- found that following cocaine administration (1 day or 5 tory synapses on VTA DA neurons elicited by a single days), spike-timing dependent LTP, like pairing-induced drug experience or stress contributes to the early neural LTP60, was absent presumably because of occlusion by the adaptations that are required for the subsequent devel- LTP that had already occurred in vivo. opment of addiction. An important question for future Occlusion Synaptic adaptations that influence LTP in the VTA work is whether this potentiation of VTA synapses also The observation that synaptic may also occur during drug withdrawal. For example, contributes significantly to relapse, which is commonly stimulation produces no expression of c-FOS (a marker for neuronal activation) triggered by either of these experiences. This might occur further LTP (or LTD) increased sharply in rats re-exposed to an environment because stronger excitatory synapses on VTA DA neurons presumably because the 71 underlying cellular mechanisms associated with withdrawal . In a related finding, brain- will change the levels or patterns of DA release in tar- have been maximally activated derived neurotrophic factor (BDNF) levels in the VTA get structures, such as the NAc, and thereby modulate by some preceding stimulus. increased during prolonged (10–15 day) drug with- DA-dependent learned associations and behaviours5–10. When LTP (or LTD) is absent, it drawal; this has been suggested to enhance the ability to It will be important to establish stronger links between is often difficult to determine elicit LTP using a weak induction protocol that normally the synaptic changes in the VTA that are triggered whether it has been ‘occluded’ 72 or blocked by inhibition or does not elicit LTP . This effect might be related to the rapidly by addictive drugs and the downstream neural inactivation of one or more ability of BDNF when injected into the VTA immediately circuit adaptations that ultimately underlie the persistent essential cellular mechanisms. after a regimen of cocaine self-administration to enhance behavioural changes that define addiction.

Ca2+-permeable AMPARs in the VTA and transiently enhanced NMDAR-mediated EPSCs Most AMPARs are heteromers of at least two different (but not AMPAR-mediated EPSCs) in VTA DA neu- AMPAR subunits including GluR2 and, therefore, are rons; this effect was blocked by the OXR1 antagonist, not very permeable to Ca2+ (REF. 74). However, AMPARs SB 334867 (REF.91) (FIG. 5). Interestingly, an increase in lacking GluR2, such as GluR1 homomeric receptors, AMPAR miniature EPSCs (mEPSCs) and the AMPAR/ are Ca2+-permeable and, therefore, can initiate Ca2+- NMDAR ratio was observed 3–4 hours later. The triggered intracellular signalling cascades in a manner delayed potentiation of AMPAR mEPSCs did not occur that is analogous to NMDARs. Based on the dramatic when orexin A was applied with an NMDAR antagonist, behavioural changes elicited by the overexpression of suggesting that the increase in NMDAR-mediated GluR1 in VTA DA neurons, increased numbers of syn- responses promotes LTP induction at these synapses aptic Ca2+-permeable AMPARs have been proposed to within a few hours. Consistent with these observations, have an important role in the development of behavioural when mice were pre-injected with SB 334867 before sensitization75. Recent electrophysiological evidence each of the daily cocaine injections for 5 days, the from cortical and hippocampal preparations suggests AMPAR/NMDAR ratio was not significantly increased that GluR2-lacking AMPARs can in fact be inserted at 24 hours after the last cocaine treatment. Moreover, potentiated synapses that previously expressed GluR2- cocaine-induced locomotor sensitization was prevented containing AMPARs76–78 (although some of these results by either systemic administration of SB 334867 or local are controversial)79,80. Similarly, insertion of GluR2- injection of SB 334867 into the VTA91. lacking AMPARs has also recently been suggested to occur at synapses on VTA DA neurons after in vivo cocaine administration. Notably, this process can be reversed by mGluR-dependent LTD, which results from the removal of GluR2-lacking synaptic AMPARs and their replacement by GluR2-containing ones59,81,82. This is a potentially important observation as it suggests a strategy for reversing or preventing the drug-induced Orexin Orexin 1 receptor LTP at VTA DA cell synapses. A limitation to this hypothesis, however, is that the well-established biophysical properties of GluR2-lacking AMPARs74 suggest that following cocaine administration, only a modest fraction of synaptic AMPARs can PKC be GluR2-lacking as such receptors show strong inward rectification and pass little current at positive membrane potentials. If most synaptic AMPARs following cocaine administration were GluR2-lacking, the increase in the AMPAR/NMDAR ratio, which has been measured at +40 mV in all studies to date (FIG. 3), would be minimal NMDAR or absent. Furthermore, other groups have not observed After 3-4 a change in the rectification of AMPAR EPSCs despite hours simultaneously observing a clear drug-induced increase in the AMPAR/NMDAR ratio62. Technical differences in experimental conditions may contribute to these incon- sistencies and, thus, this intriguing hypothesis deserves further study. Glutamate input Orexin receptors and LTP in the VTA AMPAR Orexin A and B (also known as hypocretin-1 and -2) are neuropeptides with behavioural effects on arousal and Addiction- associated feeding that are synthesized exclusively in neurons of behaviours the lateral hypothalamus83,84. Growing evidence links the orexin system with reward and reinforcement85. For example, orexin neurons are activated in response to rewarding Figure 5 | Orexin A enhances NMDAR EPSCs in VTA stimuli such as food or addictive drugs85 and innervate dopamine neurons. OrexinergicNatur afferentse Reviews from | Neur theoscienc laterale the dendrites of dopaminergic neurons86 as well as cells hypothalamus innervate dopamine neurons of the ventral tegmental area (VTA). Over a period of several minutes, within the NAc87. Furthermore, orexin A re-instates 88 activation of orexin 1 receptors by orexin A enhances cocaine-seeking in rats , whereas an orexin receptor 1 !-methyl-"-aspartate receptor (NMDAR) excitatory (OXR1) antagonist blocks acquisition of conditioned place postsynaptic currents (EPSCs) by a protein kinase C (PKC)- 89,90 preference when delivered locally to the VTA . dependent insertion of NMDARs at synapses. This process Recently, a synaptic effect of orexin A that might is followed within 3–4 hours by an increase of A-amino-3- contribute to its behavioural effects has been reported. hydroxy-5-methyl-4-isoxazole propionic acid receptor Brief (5 minute) exposure to orexin A in slices rapidly (AMPAR) EPSCs that results from AMPAR insertion91.

These data suggest that OXR1 receptors in the VTA LTPGABA is expected to increase the firing of VTA DA are necessary for the development of locomotor sensitiza- neurons following morphine exposure. Restoration of tion, possibly by increasing NMDAR-mediated currents. guanylate cyclase function might represent an avenue Cocaine application to slices, by D5-like receptor activ- that could be exploited to rescue or prevent the loss of ation, has also been found to increase NMDAR-mediated normal inhibition after opiate exposure. It will be impor- EPSCs in VTA DA cells92 suggesting that upregulation of tant to determine whether other drugs of abuse target

NMDARs through multiple mechanisms could be a criti- LTPGABA in the same manner as morphine. cal trigger for the drug-induced LTP of AMPAR-mediated responses. Indeed, previous work has shown that NMDAR Plasticity in the NAc activation is a prerequisite for psychostimulants to elicit As is the case for the VTA, it is generally accepted that locomotor sensitization as well as to increase the AMPAR/ plasticity in the NAc and its associated circuitry has a key NMDAR ratio15–17,60. Together, these findings suggest role in many forms of reward-dependent learning and that the orexin pathway to the VTA is a critical player in that the powerful reinforcing effects of drugs of abuse can reward-based learning and memory and, therefore, also hijack these circuits to produce the pathological behav- in addiction. Important questions for the future include iours that define addiction5–10,26–28,97–100. Adaptations in how cocaine promotes the release of orexin A from lateral the excitatory synaptic inputs onto NAc medium spiny hypothalamic neurons and whether other drugs of abuse neurons seem to be particularly important for mediat- also work through the orexin system. ing addiction-related changes in behaviour and, thus, the detailed molecular mechanisms by which drugs of GABAergic VTA synapses as drug targets abuse modify excitatory synaptic function in the NAc Inhibitory synapses in the VTA also have a critical role have received increased attention. The major cell type in in controlling the firing rate of DA neurons. Blockade of the NAc is the GABAergic medium spiny neuron, which

GABAA receptors strongly increases DA cell firing both makes inhibitory connections with cells in the ventral in vivo and in slices93,94, and thus plasticity of GABAergic pallidum and VTA, and receives excitatory inputs from synapses might have a profound influence on the activity limbic structures, specifically the hippocampus and of VTA DA neurons. Indeed, daily injections of cocaine amygdala, as well as the PFC (FIG. 1). The NAc can be fur-

(over 5–7 days) decreased the size of GABAA recep- ther divided into two subregions termed the core and the tor-mediated miniature inhibitory postsynaptic currents shell, which differ in their detailed histochemical archi- (IPSCs) on DA neurons as well as their maximal evoked tecture and, presumably, their function. The NAc core 56 GABAA currents . Associated with this decrease in is often considered a functional extension of the dorsal inhibition was an increased likelihood of firing to a fixed striatum and may be particularly important for instru- stimulus. Although the mechanism by which cocaine mental learning, including cue-induced reinstatement depresses GABAergic function is unknown, this work of drug seeking behaviour98. The shell is a transitional

suggested that enhancing GABAA currents might be a zone between the striatum and extended amygdala and way to counteract some of the neuroadaptations caused may be preferentially involved in mediating the primary by cocaine in the VTA. Nicotine also produces a per- reinforcing effects of drugs of abuse99,100. sistent depression of GABA release for at least an hour; longer-term effects have yet to be examined95. LTD and LTP in the NAc. Within the NAc, the phenom-

GABAA receptor synapses on VTA dopamine neu- ena of LTD and LTP have once again served as models rons also exhibit robust LTP in slices from naive animals for the types of changes that drugs of abuse may cause in response to high-frequency stimulation96 (FIG. 4). This in this brain region. However, drawing parallels between LTP of IPSCs can be triggered by NMDAR activation, LTP and LTD in the NAc and drug-induced plasticity is suggesting that NMDAR-dependent LTP at excitatory particularly difficult because very little is known about synapses may ordinarily be accompanied by NMDAR- the mechanisms of synaptic plasticity at NAc excitatory dependent LTP of inhibitory synapses, thus, helping to synapses. NMDAR-dependent LTP and LTD have been keep the firing rate of DA cells relatively constant. Several reported to occur at these synapses58,101–103 as well as lines of evidence suggest that this potentiation (termed eCB-LTD104,105. Similar to these forms of plasticity at

LTPGABA) is initiated by the postsynaptic release of nitric hippocampal CA1 pyramidal cell synapses, LTP in the oxide (NO) from DA cells that feeds back on inhibitory NAc is enhanced in mice lacking postsynaptic density terminals to activate guanylate cyclase, which in turn protein 95 (PSD-95)103, whereas NMDAR-dependent causes a long-lasting enhancement of GABA release96. LTD appears to involve the endocytosis of AMPARs106 (FIG. 2c) Importantly, LTPGABA was entirely absent in slices from . eCB-LTD involves activation of postsynaptic animals treated with morphine 24 hours earlier. This mGluRs leading to the release of an endogenous cannabi- Inhibitory postsynaptic currents was due to a disruption of the coupling between NO noid that activates presynaptic CB1 receptors to cause a 45,104 (IPSCs). Currents measured and guanylate cyclase, as in slices from morphine treated long-lasting decrease in glutamate release (FIG. 2e). using electrophysiological animals, cyclic GMP was still effective at producing Does in vivo administration of a drug of abuse recordings from a single neuron LTPGABA but NO itself was not. These data suggest that a modify excitatory synapses in the NAc by activating or while electrically stimulating single in vivo exposure to morphine either causes a loss interfering with one or more of these synaptic plasticity axons to release neurotransmitter. For the of guanylate cyclase from presynaptic GABA-releasing mechanisms? Evidence that this may occur came from 96 purposes of this Review, IPSCs terminals or renders guanylate cyclase NO-insensitive . studies, similar to those performed in the VTA, in which are GABA-mediated. Together with LTP at excitatory synapses, this loss of the AMPAR/NMDAR ratio was used as a measure of

basal synaptic strength107. Although unlike the VTA, a this increase was reversed by an in vivo challenge dose single in vivo dose of cocaine caused no change in this of cocaine. These findings are consistent not only with ratio, chronic (5 days) cocaine administration followed the behavioural effects of the LTD-blocking peptides but by 10–14 days of withdrawal caused a decrease in the also with reports that chronic in vivo administration of AMPAR/NMDAR ratio recorded from medium spiny psychostimulants increase the density of dendritic spines neurons in the NAc shell. This decrease probably involves on NAc medium spiny neurons110 , the surface expression a downregulation of AMPARs and shares mechanisms of AMPARs in the NAc111 and the behavioural responses with NMDAR-dependent LTD as this form of plasticity to infusion of AMPA into the NAc112,113. was reduced by the cocaine experience107. Analogous All of the work discussed thus far presumably deals results were found in subsequent studies examining the with a form of NMDAR-dependent LTD in the NAc synaptic effects of prolonged cocaine self-administra- that is mainly due to the loss of synaptic AMPARs3. tion. Evoked excitatory synaptic responses recorded As mentioned above, eCB-LTD also exists in the NAc extracellularly in the shell were reduced108, and in both and can be modified by in vivo administration of drugs the core and shell LTD was absent97. Importantly, animals of abuse. Specifically, a single dose of cocaine or the that self-administered food or that received cocaine pas- partial CB1 receptor agonist $9-tetrahydrocannabinol sively in a yoked design still expressed LTD. Furthermore, ($9-THC), or chronic administration of $9-THC, abol- in the core (but not the shell), LTD could not be ishes eCB-LTD105,114,115. The effects of cocaine on this elicited in cocaine self-administering animals even after form of plasticity might be due to a loss of the surface 21 days of abstinence97. expression of postsynaptic group I mGluRs (that is, These results suggest that in vivo cocaine administra- mGluR5), which are required for endocannabinoid pro- tion promotes depression of excitatory synaptic trans- duction116. Chronic $9-THC, however, appears to cause mission in the NAc (FIG. 6), although the exact circuits a functional downregulation of presynaptic CB1 recep- in which this occurs (for example, core versus shell) and tors105,116 and a homeostatic upregulation of a different how long this synaptic plasticity lasts might depend on form of presynaptic LTD that is triggered by activation additional variables, such as whether the drug was self- of presynaptic group II mGluRs116. administered or not. The approach used in these stud- Compared with the effects of drugs of abuse on LTD, ies, however, cannot address the behavioural relevance much less is known about their influence on LTP in the of such synaptic modifications. To directly examine NAc. In one study, chronic (5 day) cocaine administra- whether LTD in the NAc has a role in amphetamine- tion led to an enhancement of LTP in slices of the NAc induced behavioural sensitization, a membrane perme- measured 2–3 days after the last injection103. This was able peptide that was shown to block clathrin-mediated associated with a decrease in the NAc levels of the syn- endocytosis, and thus LTD, was administered into the aptic scaffolding protein PSD-95 and, consistent with NAc of sensitized rats immediately before a challenge this observation, LTP in the NAc was also enhanced in dose of amphetamine106. Remarkably, this prevented mutant mice lacking fully functional PSD-95 (REF.103). the increase in locomotor activity normally elicited by the However, LTP in the NAc in vivo elicited by tetanization drug and, therefore, the expression of behavioural sensit- of hippocampal afferents was reported to be blocked, ization, while having no effect on the acute response to not enhanced, 18–25 days after chronic (6 day) cocaine amphetamine in non-sensitized animals. Injection of administration117. The interpretation of the results from this peptide into the VTA, however, had no behavioural these studies is limited by the fact that extracellular effect. recording techniques were used, making it difficult to Although these behavioural results provide support for assess what proportion of any observed changes in the the functional importance of the drug-induced decreases responses was due to synaptic modifications rather than in excitatory synaptic strength in the NAc reported in other parameters such as cell excitability. earlier studies, an important caveat is that the peptide blocked the acute expression of behavioural sensitization Other approaches examining NAc synaptic plasticity. (its effects on the development of sensitization were not Although examining drug-induced changes tested)106. This observation suggests that acute ampheta- in synaptic function, by definition, requires measuring mine administration elicits LTD in the NAc rather than synaptic responses using electrophysiological record- LTD induced by the previous drug experience. Indeed, ing techniques, additional assays have provided useful the original observation of a decrease in the AMPAR/ information about the potential role of synaptic adapta- NMDAR ratio in the NAc of cocaine-sensitized animals tions in the NAc in addiction. Extracellular glutamate was made following a challenge dose of cocaine107. To levels in the NAc decrease during cocaine withdrawal, determine whether this single challenge dose of cocaine partly because of a downregulation of the glial cystine- 99,118 Yoked design influenced excitatory synaptic strength in sensitized ani- glutamate transporter . Specific isoforms of the Experimental protocol in which mals, AMPAR/NMDAR ratios were measured in NAc synaptic protein Homer, which helps cluster mGluRs a ‘yoked’ control animal shell neurons from animals who received a repeated and associated signalling proteins, are also decreased receives a drug administered cocaine treatment that was sufficient to cause sensitization by cocaine administration119 as is PSD-95103. All of by the investigator in a non- but did not receive a single dose of cocaine 10–14 days these changes have been shown to affect drug-induced contingent manner, in the 109 same amount and temporal later . Surprisingly, the AMPAR/NMDAR ratio was behavioural adaptations but exactly how they influence pattern as an animal that is significantly increased, probably due to an upregulation excitatory synaptic function and plasticity within the self-administering the drug. of AMPARs and, consistent with the previous study, NAc needs further investigation. A different approach

a Control b Cocaine c Cocaine withdrawal

NMDAR

AMPAR

Glu

eCB eCB eCB

CB1R

mGluR

AMPARs eCB-LTD AMPARs Spines decreases release of Glu eCB-LTD eCB-LTD Figure 6 | Drugs of abuse modulate synaptic function and plasticity in the nucleus accumbens (NAc). Highly simplified Nature Reviews | Neuroscience summary of the effects of cocaine (and perhaps other drugs of abuse) on synaptic function and plasticity in the NAc. a | In control conditions, a normal complement of synaptic A-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors (AMPARs) and !-methyl-"-aspartate receptors (NMDARs) exists, and endocannabinoid-mediated long-term depression (eCB-LTD) can be induced. b | Cocaine administration causes a loss of synaptic AMPARs due to their internalization106,107 and a concomitant block of eCB-LTD114. c | During withdrawal from chronic cocaine administration there is an increase in surface AMPARs and in dendritic spine density109–111, perhaps due to homeostatic synaptic scaling47 in response to the decrease in synaptic strength and excitability elicited by cocaine (seen in b). The synaptic adaptations during withdrawal may differ between the NAc shell and core (not shown) as LTD is blocked in the core, but not the shell, 21 days after withdrawal from cocaine self-administration97. The effects of withdrawal from cocaine administration on LTP are also complex (not shown)103,117. CB1R, cannabinoid receptor 1; Glu, glutamate; mGluR, metabotropic glutamate receptor.

has been to examine the behavioural consequences Other cellular adaptations. Although the focus of this of overexpression of wild type or dominant nega- review is on synaptic plasticity, it is important to men- tive AMPAR subunits in the NAc. Overexpression of tion that drugs of abuse may cause cellular adaptations GluR1 in the NAc, which would be expected to increase that do not directly affect synaptic transmission, yet may excitatory synaptic strength, facilitated the extinction profoundly influence neural circuit function. Among of cocaine-seeking responses120, attenuated the reward- the most important of these non-synaptic adaptations ing effects of cocaine in conditioned place preference121 are changes in the intrinsic excitability of NAc neurons. and increased brain reward stimulation thresholds122. Chronic administration of psychostimulants causes a Conversely, overexpression of a pore-dead GluR1, decrease in neuronal excitability in the NAc, due to the which should decrease synaptic strength, exacerbated modulation of voltage-dependent conductances123–125, cocaine-induced locomotor sensitization and rein- and this adaptation alone is sufficient to cause a sensi- statement of drug seeking (R.K. Bachtell and D.W. Self, tized behavioural response to cocaine125. Another personal communication). These results are consistent important issue that is just beginning to be explored125 with the idea that rapid drug-induced depression of is how drug-induced gene transcription and epigenetic excitatory synaptic function in the NAc contributes to regulation126 influence synaptic and circuit function in the long-lasting sensitized behavioural responses caused the mesolimbic dopamine system and, thereby, exert by previous drug exposure. their behavioural effects.

BNST and amygdala Conclusions Although the VTA–NAc axis has been the most exten- We have attempted to review the growing literature on sively studied circuit in relation to motivation and the synaptic modifications elicited by drugs of abuse in reward in drug addiction, it is clear that other brain the mesolimbic DA system and how these may contribute regions are essential components as well. Here we briefly to the development of addiction. The changes that occur review evidence that synaptic plasticity in two additional are complex and much work remains to establish which regions, the bed nucleus of the stria terminalis (BNST) of the myriad reported changes in synaptic function are and amygdala, might also be modified by drugs of abuse. the most behaviourally relevant. Nonetheless, it can be The BNST is considered a component of the extended argued that more than any other commonly studied form amygdala and has a role in stress- and reward-related of experience-dependent plasticity, we are beginning to limbic circuitry. It contributes to stress-induced relapse understand the potential causal relationships between to drug-seeking behaviour and its neurons project to the neural circuit adaptations elicited by drugs of abuse areas involved in feeding and reward processing, includ- and the behavioural consequences of that experience. ing the NAc and VTA127,128. NMDAR-dependent LTP can To determine whether these synaptic adaptations are be triggered in the BNST and acute ethanol impairs this important in the development or maintenance of human LTP in part by attenuating NMDAR-mediated synaptic addictions, however, will require much further effort. currents129. In contrast, potentiation of the AMPAR/ By necessity the study of synaptic plasticity depends on NMDAR ratio at excitatory synapses in the ventral lat- the use of electrophysiological recording techniques, eral BNST occurred in rats that self-administered either which permit real-time measurements of the functional palatable food pellets or intravenous cocaine daily for state of synapses in defined populations of neurons. eleven days130. Intriguingly, neither experimenter- Because recordings from individual cells are difficult administered food nor drug produced an enhancement to make in older animals, many of the studies to date of the ratio, suggesting that LTP at this set of excitatory have been carried out in young animals. Furthermore, synapses in the BNST occurs only when an operant task most studies have not used protocols involving drug self- is performed to obtain a reward. LTD in the BNST has administration (in part because of the necessity of using also been examined and requires activation of group I older animals), which is the most faithful representation mGluRs, which elicit LTD via an eCB-independ- of human drug abuse. Thus, in future work, it will be ent, extracellular-signal-regulated kinase 1 (ERK1)- important to balance the pragmatic advantages of using dependent mechanism131. This form of LTD is prevented simple in vivo procedures involving direct administra- by chronic (10 day) but not single day administration of tion of drugs by the experimenter with more difficult, cocaine. Further work will be necessary to determine but clinically relevant, protocols that more accurately whether this effect of cocaine is due to an occlusion or mirror human drug-seeking and drug-taking behaviour. block of LTD. Whole cell recording from individual cells continues The amygdala is known to be involved in forms of to provide the most sensitive and complete picture of memory that involve a strong emotional component and synaptic function. However, combining this technique LTP in the lateral nucleus of the amgydala has long been with fluorescence imaging techniques or with simpler suggested to have an important role in fear condition- electrophysiological approaches, such as field potential ing132. In human subjects, cocaine-associated cues that recordings that can more easily be made in vivo, may induce craving in substance abusers robustly alter neuro- improve our ability to assay synaptic function across nal activity in the amygdala133. Recently, LTP at excitatory larger cell populations. synapses in the central amygdala was found to increase The cell populations in critical brain regions such as dramatically two weeks after withdrawal from chronic the VTA and NAc are likely to be heterogeneous, dif- (14 day) cocaine administration134. However, this effect fering in their molecular and physiological properties was not seen 24 hours after the last cocaine injection, as well as their anatomical connectivity. Therefore, it indicating that it increases with time of withdrawal. Both will be essential to determine whether drugs of abuse NMDARs and corticotropin-releasing factor (CRF) are influence the different cell types in the same manner. essential for the LTP to be triggered in vitro. This latter Compounding this issue is the difficulty of knowing observation is of particular interest because CRF release which specific sets of synapses (that is, inputs) on a in the amygdala is increased during acute withdrawal135. given cell are being assayed and modified by the drug Even a brief application of CRF was reported to potenti- experience. The powerful combination of single cell ate excitatory synaptic currents and this effect was also electrophysiology with animals that are genetically modi- enhanced after 2 weeks withdrawal from cocaine136. fied, by engineering from birth or by using viral vectors Enhanced synaptic plasticity in the central amygdala in the postnatal brain, offers an excellent approach to may be one cellular adaptation contributing to the CRF- define the precise neuronal populations and physi- dependent signalling that appears to cause anxiety and ological parameters modified by exposure to drugs of stress responses during withdrawal from cocaine. If the abuse. A recent exciting advance is the use of genetically ability to elicit LTP in the amygdala does in fact correlate modified mice in which fluorescent markers (for exam- with the aversiveness of withdrawal symptoms, it would ple, green flourescent protein (GFP)) are expressed in be of interest to see whether LTP returns to control specific, molecularly-identified cell populations137. Such levels after an additional drug exposure that reduces mice have already facilitated our understanding of the withdrawal symptoms. cell type-specific synaptic modifications that occur in

the dorsal striatum following dopamine depletion138,139 behaviour in unanaesthetized animals during neuronal as well as cocaine-induced changes in spine density activation or inactivation142. With further advances in our in the NAc140. Additional molecular approaches, such as understanding of the molecular mechanisms of synaptic the restricted expression of light-activated ion channels plasticity in the key circuits mediating addiction com- and pumps that depolarize or hyperpolarize neurons141, bined with more sophisticated applications of animal and should allow activation or inhibition of specific affer- human brain imaging, we anticipate that our knowledge ent inputs and, thereby, permit a more precise assay of of the neural circuit adaptations that underlie addiction synaptic adaptations in specific mesolimbic DA circuits. will grow exponentially. These advances will facilitate This approach could potentially also be exploited in more informed and effective approaches to the treatment experiments in vivo by implanting a fibre optic light of addiction and related disorders, which remain one of source within a relevant brain structure and monitoring society’s most challenging health problems.

1. Ramón y Cajal, S. La fine structure des centres 20. Kim, H. S., Park, W. K., Jang, C. G. & Oh, S. Inhibition 41. Morishita, W., Marie, H. & Malenka, R. C. Distinct nerveux. Proc. R. Soc. Lond. 55, 444–468 (1894). by MK-801 of cocaine-induced sensitization, triggering and expression mechanisms underlie LTD of 2. Bliss, T. V. P. & Lomo, T. Long-lasting potentiation of conditioned place preference, and dopamine-receptor AMPA and NMDA synaptic responses. Nature synaptic transmission in the dentate area of the supersensitivity in mice. Brain. Res. Bull. 40, Neurosci. 8, 1043–1050 (2005). anaesthetized rabbit following stimulation of the 201–207 (1996). 42. Ito, M. Long-term depression. Annu. Rev. Neurosci. perforant path. J. Physiol. 232, 331–356 (1973). 21. Tzschentke, T. M. & Schmidt, W. J. N-methyl-D-aspartic 12, 85–102 (1989). The initial, now classic, description of LTP in the acid-receptor antagonists block morphine-induced 43. Pfeiffer, B. E. & Huber, K. M. Current advances in local hippocampus. conditioned place preference in rats. Neurosci. Lett. protein synthesis and synaptic plasticity. J. Neurosci. 3. Malenka, R. C. & Bear, M. F. LTP and LTD: an 193, 37–40 (1995). 26, 7147–7150 (2006). embarrassment of riches. Neuron 44, 5–21 (2004). 22. Kalivas, P. W. & Stewart, J. Dopamine transmission in 44. Wilson, R. I. & Nicoll, R. A. Endocannabinoid signaling 4. Foeller, E. & Feldman, D. E. Synaptic basis for the initiation and expression of drug- and stress- in the brain. Science 296, 678–682 (2002). developmental plasticity in somatosensory cortex. induced sensitization of motor activity. Brain Res. Rev. 45. Chevaleyre, V., Takahashi, K. A. & Castillo, P. E. Curr. Opin. Neurobiol. 14, 89–95 (2004). 16, 223–244 (1991). Endocannabinoid-mediated synaptic plasticity in the 5. Hyman, S. E. & Malenka, R. C. Addiction and the brain: 23. Robinson, T. E. & Berridge, K. C. The neural basis of CNS. Annu. Rev. Neurosci. 29, 37–76 (2006). the neurobiology of compulsion and its persistence. drug craving: an incentive-sensitization theory of 46. Chevaleyre, V. et al. Endocannabinoid-mediated long- Nature Rev. Neurosci. 2, 695–703 (2001). addiction. Brain Res. Rev. 18, 247–291 (1993). term plasticity requires cAMP/PKA signaling and 6. Kalivas, P. W. & Volkow, N. D. The neural basis of 24. To n g , Z . Y., Overton, P. G. & Clark, D. Chronic RIM1A. Neuron 54, 801–812 (2007). addiction: a pathology of motivation and choice. administration of (+)-amphetamine alters the 47. Tu r r i g i a n o , G . G. & Nelson, S. B. Homeostatic Am. J. Psychiatry 162, 1403–1413 (2005). reactivity of midbrain dopaminergic neurons to plasticity in the developing nervous system. Nature 7. Montague, P. R., Hyman, S. E. & Cohen, J. D. prefrontal cortex stimulation in the rat. Brain Res. Rev. Neurosci. 5, 97–107 (2004). Computational roles for dopamine in behavioural 674, 63–74 (1995). 48. Wierenga, C. J., Ibata, K. & Turrigiano, G. G. control. Nature 431, 760–767 (2004). 25. Wolf, M. E. The role of excitatory amino acids in Postsynaptic expression of homeostatic plasticity at 8. Hyman, S. E., Malenka, R. C. & Nestler, E. J. Neural behavioral sensitization to psychomotor stimulants. neocortical synapses. J. Neurosci. 25, 2895–2905 mechanisms of addiction: the role of reward–related Prog. Neurobiology 54, 1–42 (1998). (2005). learning and memory. Annu. Rev. Neurosci. 29, 26. Everitt, B. J. & Wolf, M. E. Psychomotor stimulant 49. Stellwagen, D. & Malenka, R. C. Synaptic scaling 565–598 (2006). addiction: a neural systems perspective. J. Neurosci. mediated by glial TNF-A. Nature 440, 1054–1059 9. Kauer, J. A. Learning mechanisms in addiction: 22, 3312–3320 (2002). (2006). synaptic plasticity in the ventral tegmental area as a 27. Kalivas, P. W. Glutamate systems in cocaine addiction. 50. Di Chiara, G. & Imperato, A. Drugs abused by humans result of exposure to drugs of abuse. Annu. Rev. Curr. Opin. Pharmacol. 4, 23–29 (2004). preferentially increase synaptic dopamine Physiol. 66, 447–475 (2004). 28. Baler, R. D. & Volkow, N. D. Drug addiction: the concentrations in the mesolimbic system of freely 10. Kelley, A. E. Memory and addiction: shared neural neurobiology of disrupted self-control. Trends Mol. moving rats. Proc. Natl Acad. Sci. USA 85, circuitry and molecular mechanisms. Neuron 44, Med. 12, 559–566 (2006). 5274–5278 (1980). 161–179 (2004). 29. Malenka, R. C. & Nicoll, R. A. Long-term potentiation 51. Omelchenko, N. & Sesack, S. R. Glutamate synaptic 11. Badiani, A. & Robinson, T. E. Drug-induced — a decade of progress? Science 285, 1870–1874 inputs to ventral tegmental area neurons in the rat neurobehavioral plasticity: the role of environmental (1999). derive primarily from subcortical sources. context. Behav. Pharmacol. 15, 327–339 (2004). 30. Lynch, M. A. Long-term potentiation and memory. Neuroscience 146, 1259–1274 (2007). 12. Kitamura, O., Wee, S., Specio, S. E., Koob, G. F. & Physiol. Rev. 84, 87–136 (2004). 52. Carr, D. B. & Sesack, S. R. Projections from the rat Pulvirenti, L. Escalation of methamphetamine self- 31. Yuste, R. & Bonhoeffer, T. Morphological changes in prefrontal cortex to the ventral tegmental area: target administration in rats: a dose-effect function. dendritic spines associated with long-term synaptic specificity in the synaptic associations with Psychopharmacology (Berl) 186, 48–53 (2006). plasticity. Annu. Rev. Neurosci. 24, 1071–1089 (2001). mesoaccumbens and mesocortical neurons. 13. Morris, R. G. Elements of a neurobiological theory of 32. Matsuzaki, M., Honkura, N., Ellis-Davies, G. C. & J. Neurosci. 20, 3864–3873 (2000). hippocampal function: the role of synaptic plasticity, Kasai, H. Structural basis of long-term potentiation in 53. Overton, P. G., Richards, C. D., Berry, M. S. & Clark, D. synaptic tagging and schemas. Eur. J. Neurosci. 23, single dendritic spines. Nature 429, 761–766 (2004). Long-term potentiation at excitatory amino acid 2829–2846 (2006). 33. Lau, C. G. & Zukin, R. S. (2007). NMDA receptor synapses on midbrain dopamine neurons. Neuroreport 14. Schenk, S., Valadez, A., Worley, C. M. & McNamara, C. trafficking in synaptic plasticity and neuropsychiatric 10, 221–226 (1999). Blockade of the acquisition of cocaine self- disorders. Nature Rev. Neurosci. 8, 413–426. 54. Bonci, A. & Malenka, R. C. Properties and plasticity of administration by the NMDA antagonist MK-801 34. Nicoll, R. A. & Schmitz, D. Synaptic plasticity at excitatory synapses on dopaminergic and GABAergic (dizocilpine). Behav. Pharmacol. 4, 652–659 hippocampal mossy fibre synapses. Nature Rev. cells in the ventral tegmental area. J. Neurosci. 19, (1993). Neurosci. 6, 863–876 (2005). 3723–3730 (1999). 15. Kalivas, P. W. & Alesdatter, J. E. Involvement of 35. Yeckel, M. F., Kapur, A. & Johnston, D. Multiple forms 55. Mansvelder, H. D. & McGehee, D. S. Long-term NMDA receptor stimulation in the ventral tegmental of LTP in hippocampal CA3 neurons use a common potentiation of excitatory inputs to brain reward areas area and amygdala in behavioral sensitization to postsynaptic mechanism. Nature Neurosci. 2, by nicotine. Neuron 27, 349–357 (2000). cocaine. 625–633 (1999). 56. Liu, Q. S., Pu, L. & Poo, M. M. Repeated cocaine J. Pharmacol. Exp. Ther. 267, 486–495 (1993). 36. Contractor, A., Rogers, C., Maron, C., Henkemeyer, M., exposure in vivo facilitates LTP induction in midbrain 16. Harris, G. C., Wimmer, M., Byrne, R. & Aston-Jones, G. Swanson, G. T. & H e i n e m a n n S . F. Trans-synaptic Eph dopamine neurons. Nature 437, 1027–1031 Glutamate-associated plasticity in the ventral receptor-ephrin signaling in hippocampal mossy fiber (2005). tegmental area is necessary for conditioning LTP. Science 296, 1864–1869 (2002). 57. Jones, S., Kornblum, J. L. & Kauer, J. A. Amphetamine environmental stimuli with morphine. Neuroscience 37. Castillo, P. E., Schoch, S., Schmitz, F., Sudhof, T. C. & blocks long-term synaptic depression in the ventral 129, 841–847 (2004). Malenka, R. C. RIM1A is required for presynaptic long- tegmental area. J. Neurosci. 20, 5575–5580 17. Harris, G. C. & Aston-Jones, G. Critical role for ventral term potentiation. Nature 415, 327–330 (2002). (2000). tegmental glutamate in preference for a cocaine- 38. Castillo, P. E. et al. Rab3A is essential for mossy fibre 58. Thomas, M. T. , M a l e n ka , R . C. & Bonci, A. Modulation conditioned environment. Neuropsychopharmacology long-term potentiation in the hippocampus. Nature of long-term depression by dopamine in the mesolimbic 28, 73–76 (2003). 388, 590–593 (1997). system. J. Neurosci. 20, 5581–5586 (2000). 18. Karler, R., Calder, L. D., Chaudhry, I. A. & Turkanis, S. A. 39. Carroll, R. C., Beattie, E. C., von Zastrow, M. & 59. Bellone, C. & Luscher, C. mGluRs induce a long-term Blockade of “reverse tolerance” to cocaine and Malenka, R. C. Role of AMPA receptor endocytosis in depression in the ventral tegmental area that involves amphetamine by MK-801. Life Sciences 45, 599–606 synaptic plasticity. Nature Rev. Neurosci. 2, 315–324 a switch of the subunit composition of AMPA (1989). (2001). receptors. Eur. J. Neurosci. 21, 1280–1288 (2005). 19. Jeziorski, M., White, F. J. & Wolf, M. E. MK-801 40. Selig, D. K., Hjelmstad, G. O., Herron, C., Nicoll, R. A. 60. Ungless, M. A., Whistler, J. L., Malenka, R. C. & prevents the development of behavioral sensitization & Malenka, R. C. Independent mechanisms for long- Bonci, A. Single cocaine exposure in vivo induces long- during repeated morphine administration. Synapse term depression of AMPA and NMDA responses. term potentiation in dopamine neurons. Nature 411, 16, 137–147 (1994). Neuron 15, 417–426 (1995). 583–587 (2001).

61. Saal, D., Dong, Y., Bonci, A. & Malenka, R. C. Drugs of References 81 and 82 present evidence that a 101. Kombian, S. B. & Malenka, R. C. Simultaneous LTP of abuse and stress trigger a common synaptic novel form of mGluR-LTD reverses the cocaine- non-NMDA- and LTD of NMDA-receptor-mediated adaptation in dopamine neurons. Neuron 37, induced LTP at excitatory synapses on ventral responses in the nucleus accumbens. Nature 368, 577–582 (2003). tegmental area dopamine cells. 242–246 (1994). 62. Faleiro, L. J., Jones, S. & Kauer, J. A. Rapid synaptic 83. de Lecea, L. et al. The hypocretins: hypothalamus- 102. Schramm, N. L., Egli, R. E. & Winder, D. G. LTP in the plasticity of glutamatergic synapses on dopamine specific peptides with neuroexcitatory activity. mouse nucleus accumbens is developmentally neurons in the ventral tegmental area in response to Proc. Natl Acad. Sci. USA 95, 322–327 (1998). regulated. Synapse 45, 213–219 (2002). acute amphetamine injection. 84. Sakurai, T. et al. Orexins and orexin receptors: a 103. Yao, W. D. et al. Identification of PSD-95 as a Neuropsychopharmacology 29, 2115–2125 (2004). family of hypothalamic neuropeptides and G protein- regulator of dopamine-mediated synaptic and References 60–62 demonstrate that in vivo coupled receptors that regulate feeding behavior. behavioral plasticity. Neuron 41, 625–638 (2004). administration of different classes of drugs of Cell 92, 573–585 (1998). 104. Robbe, D., Kopf, M., Remaury, A., Bockaert, J. & abuse, as well as acute stress, elicit LTP at 85. Harris, G. C. & Aston-Jones, G. Arousal and reward: a Manzoni, O. J. Endogenous cannabinoids mediate excitatory synapses on midbrain dopamine neurons. dichotomy in orexin function. Tre n d s N e u ro s c i . 29, long-term synaptic depression in the nucleus 63. Marinelli, M. & Piazza, P. V. Interaction between 571–577 (2006). accumbens. Proc. Natl Acad. Sci. USA 99, glucocorticoid hormones, stress and psychostimulant 86. Fadel, J. & Deutch, A. Y. Anatomical substrates of 8384–8388 (2002). drugs. Eur. J. Neurosci. 16, 387–394 (2002). orexin-dopamine interactions: lateral hypothalamic 105. Hoffman, A. F., Oz, M., Caulder, T. & Lupica, C. R. 64. Wallace, B. C. Psychological and environmental projections to the ventral tegmental area. Functional tolerance and blockade of long-term determinants of relapse in crack cocaine smokers. Neuroscience 111, 379–387 (2002). depression at synapses in the nucleus accumbens J. Subst. Abuse Treat. 6, 95–106 (1989). 87. Baldo, B. A., Daniel, R. A., Berridge, C. W. & after chronic cannabinoid exposure. J. Neurosci. 23, 65. Stewart, J. Stress and relapse to drug seeking: studies Kelley, A. E. Overlapping distributions of orexin/ 4815–4820 (2003). in laboratory animals shed light on mechanisms and hypocretin- and dopamine-B-hydroxylase 106. Brebner, K. et al. Nucleus accumbens long-term sources of long-term vulnerability. Am. J. Addict. 12, immunoreactive fibers in rat brain regions mediating depression and the expression of behavioral 1–17 (2003). arousal, motivation, and stress. J. Comp. Neurol. sensitization. Science 310, 1340–1343 (2005). 66. Piazza, P. V. & Le Moal, M. The role of stress in drug 464, 220–237 (2003). 107. Thomas, M. J., Beurrier, C., Bonci, A. & Malenka, R. C. self-administration. Trends Pharmacol. Sci. 19, 67–74 88. Boutrel, B. Hypocretins: between desire and needs. Long-term depression in the nucleus accumbens: a (1998). toward the understanding of a new hypothalamic brain neural correlate of behavioral sensitization to cocaine. 67. Dong, Y. et al. Cocaine-induced potentiation of pathway involved in motivation and addiction. Med. Nature Neurosci. 4, 1217–1223 (2001). synaptic strength in dopamine neurons: behavioral Sci. (Paris) 22, 573–575 (2006). References 106 and 107 demonstrate that in correlates in GluRA(–/–) mice. Proc. Natl Acad. Sci. 89. Harris, G. C., Wimmer, M. & Aston-Jones, G. A role for animals, which had previously been exposed to USA 101, 14282–14287 (2004). lateral hypothalamic orexin neurons in reward cocaine to elicit sensitization, a single subsequent 68. Malinow, R. & Malenka, R. C. AMPA receptor seeking. Nature 437, 556–559 (2005). dose of cocaine elicits LTD at excitatory synapses in trafficking and synaptic plasticity. Annu. Rev. Neurosci. Demonstration that orexin neurons in the lateral the nucleus accumbens and that preventing this 25, 103–126 (2002). hypothalamus play a key role in the reinstatement LTD in vivo prevents the expression of behavioral 69. Carlezon Jr, W. A. et al. Sensitization to morphine of drug-seeking behaviour at least in part due to sensitization. induced by viral-mediated gene transfer. Science 277, actions of orexin A in the ventral tegmental area. 108. Schramm-Sapyta, N. L., Olsen, C. M. & Winder, D. G. 812–814 (1997). 90. Narita, M. et al. Direct involvement of orexinergic Cocaine self-administration reduces excitatory Demonstration that viral-mediated expression of systems in the activation of the mesolimbic dopamine responses in the mouse nucleus accumbens shell. GluR1 in the ventral tegmental area enhances the pathway and related behaviors induced by morphine. Neuropsychopharmacology 31, 1444–1451 locomotor stimulatory and rewarding actions of J. Neurosci. 26, 398–405 (2006). (2006). morphine. 91. Borgland, S. L., Taha, S. A., Sarti, F., Fields, H. L. & 109. Kourrich, S., Rothwell, P., Klug, J. & Thomas, M. 70. Borgland, S. L., Malenka, R. C. & Bonci, A. Acute and Bonci, A. Orexin A in the VTA is critical for the Cocaine experience controls bidirectional synaptic chronic cocaine-induced potentiation of synaptic induction of synaptic plasticity and behavioral plasticity in the nucleus accumbens. J. Neurosci. 27, strength in the ventral tegmental area: sensitization to cocaine. Neuron 49, 589–601 7921–7928 (2007). electrophysiological and behavioral correlates in (2006). 110. Robinson, T. E. & Kolb, B. Structural plasticity individual rats. J. Neurosci. 24, 7482–7490 (2004). Demonstration that orexin A enhances NMDAR- associated with exposure to drugs of abuse. 71. Neisewander, J. L. et al. Fos protein expression and mediated synaptic currents in ventral tegmental Neuropharmacology 47, S33–S46 (2004). cocaine-seeking behavior in rats after exposure to a area (VTA) dopamine neurons and that its actions 111. Boudreau, A. C. & Wolf, M. E. Behavioral sensitization cocaine self-administration environment. J. Neurosci. in the VTA are required for behavioural to cocaine is associated with increased AMPA receptor 20, 798–805 (2000). sensitization to cocaine. surface expression in the nucleus accumbens. 72. Pu, L., Liu, Q. S. & Poo, M. M. BDNF-dependent 92. Schilstrom, B. et al. Cocaine enhances NMDA J. Neurosci. 25, 9144–9151 (2005). synaptic sensitization in midbrain dopamine neurons receptor-mediated currents in ventral tegmental area 112. Pierce, R. C., Bell, K., Duffy, P. & Kalivas, P. W. after cocaine withdrawal. Nature Neurosci. 9, cells via dopamine D5 receptor-dependent Repeated cocaine augments excitatory amino acid 605–607 (2006). redistribution of NMDA receptors. J. Neurosci. 26, transmission in the nucleus accumbens only in rats 73. Lu, L., Dempsey, J., Liu, S. Y., Bossert, J. M. & 8549–8558 (2006). having developed behavioral sensitization. Shaham, Y. A single infusion of brain-derived 93. Yim, C. Y. & Mogenson, G. J. Electrophysiological J. Neurosci. 16, 1550–1560 (1996). neurotrophic factor into the ventral tegmental area studies of neurons in the ventral tegmental area of 113. Cornish, J. L. & Kalivas, P. W. Glutamate transmission induces long-lasting potentiation of cocaine seeking Ts a i . Brain Res. 181, 301–313 (1980). in the nucleus accumbens mediates relapse in cocaine after withdrawal. J. Neurosci. 24, 1604–1611 (2004). 94. Johnson, S. W. & North, R. A. Opioids excite addiction. J. Neurosci. 20, RC89 (2000). 74. Liu, S. J. & Zukin, R. S. Ca2+-permeable AMPA dopamine neurons by hyperpolarization of local 114. Fourgeaud, L. et al. A single in vivo exposure to receptors in synaptic plasticity and neuronal death. interneurons. J. Neurosci. 12, 483–488 (1992). cocaine abolishes endocannabinoid-mediated long- Tre n d s N e u ro s c i . 30, 126–134 (2007). 95. Mansvelder, H. D., Keath, J. R. & McGehee, D. S. term depression in the nucleus accumbens. 75. Carlezon Jr, W. A. & Nestler, E. J. Elevated levels of Synaptic mechanisms underlie nicotine-induced J. Neurosci. 24, 6939–6945 (2004). GluR1 in the midbrain: a trigger for sensitization to excitability of brain reward areas. Neuron 33, 115. Mato, S. et al. A single in vivo exposure to D9THC drugs of abuse? Tre n d s N e u ro s c i . 25, 610–615 905–919 (2002). blocks endocannabinoid-mediated synaptic plasticity. (2002). 96. Nugent, F. S., Penick, E. C. & Kauer, J. A. Opioids Nature Neurosci. 7, 585–596 (2004). 76. Ju, W. et al. Activity-dependent regulation of dendritic block long-term potentiation of inhibitory synapses. References 114 and 115 demonstrate that a single synthesis and trafficking of AMPA receptors. Nature Nature 446, 1086–1090 (2007). in vivo dose of cocaine or THC abolishes Neurosci. 7, 244–253 (2004). Demonstration of LTP of inhibitory synapses on endocannabinoid-mediated LTD in the nucleus 77. Clem, R. L. & Barth, A. Pathway-specific trafficking of ventral tegmental area dopamine neurons due to a accumbens. native AMPARs by in vivo experience. Neuron 49, long-lasting enhancement of GABA release 116. Mato, S., Robbe, D., Puente, N., Grandes, P. & 663–670 (2006). triggered by NMDAR-dependent release of nitric Manzoni, O. J. Presynaptic homeostatic plasticity 78. Plant, K. et al. Transient incorporation of native oxide from the dopamine neurons. Exposure to rescues long-term depression after chronic GluR2-lacking AMPA receptors during hippocampal morphine in vivo blocks this LTP by interrupting D9-tetrahydrocannabinol exposure. J. Neurosci. 25, long-term potentiation. Nature Neurosci. 9, 602–604 the signalling from nitric oxide to guanylate 11619–11627 (2005). (2006). cyclase. 117. Goto, Y. & Grace, A. A. Dopamine-dependent 79. Adesnik, H. & Nicoll, R. A. Conservation of glutamate 97. Martin, M., Chen, B. T. , H o p f , F. W., Bowers, M. S. & interactions between limbic and prefrontal cortical receptor 2-containing AMPA receptors during long- Bonci, A. Cocaine self-administration selectively plasticity in the nucleus accumbens: disruption by term potentiation. J. Neurosci. 27, 4598–4602 abolishes LTD in the core of the nucleus accumbens. cocaine sensitization. Neuron 47, 255–266 (2007). Nature Neurosci. 9, 868–869 (2006). (2005). 80. Bagal, A. A., Kao, J. P. , Ta n g , C . M. & Thompson, S. M. 98. Cardinal, R. N. & Everitt, B. J. Neural and 118. Baker, D. A. et al. Neuroadaptations in cystine- Long-term potentiation of exogenous glutamate psychological mechanisms underlying appetitive glutamate exchange underlie cocaine relapse. Nature responses at single dendritic spines. Proc. Natl Acad. learning: links to drug addiction. Curr. Opin. Neurosci. 6, 743–749 (2003). Sci. USA 102, 14434–14439 (2005). Neurobiol. 14, 156–162 (2004). 119. Szumlinski, K. K., Kalivas, P. W. & Worley, P. F. Homer 81. Bellone, C. & Luscher, C. Cocaine triggered AMPA 99. Kalivas, P. W., Volkow, N. & Seamans, J. proteins: implications for neuropsychiatric disorders. receptor redistribution is reversed in vivo by mGluR- Unmanageable motivation in addiction: a pathology in Curr. Opin. Neurobiol. 16, 251–257 (2006). dependent long-term depression. Nature Neurosci. 9, prefrontal-accumbens glutamate transmission. Neuron 120. Sutton, M. A. et al. Extinction-induced upregulation in 636–641 (2006). 45, 647–650 (2005). AMPA receptors reduces cocaine-seeking behaviour. 82. Mameli, M., Balland, B., Lujan, R. & Luscher, C. 100. Pierce, R. C. & Kumaresan, V. The mesolimbic Nature 421, 70–75 (2003). Rapid synthesis and synaptic insertion of GluR2 for dopamine system: the final common pathway for the 121. Kelz, M. B. et al. Expression of the transcription factor mGluR-LTD in the ventral tegmental area. Science reinforcing effect of drugs of abuse? Neurosci. DFosB in the brain controls sensitivity to cocaine. 317, 530–533 (2007). Biobehav. Rev. 30, 215–238 (2006). Nature 401, 272–276 (1999).

122. To d t e n ko p f , M . S. et al. Brain reward regulated by the dorsolateral bed nucleus of the stria terminalis. 137. Gong, S. et al. A gene expression atlas of the central AMPA receptor subunits in nucleus accumbens shell. J. Neurosci. 24, 5741–5747 (2004). nervous system based on bacterial artificial J. Neurosci. 26, 11665–11669 (2006). 130. Dumont, E. C., Mark, G. P., Mader, S. & Williams, J. T. chromosomes. Nature 425, 917–925 (2003). References 120–122 demonstrate that expression Self-administration enhances excitatory synaptic 138. Kreitzer, A. C. & Malenka, R. C. Endocannabinoid- of AMPA receptor subunits in the nucleus transmission in the bed nucleus of the stria terminalis. mediated rescue of striatal LTD and motor deficits in accumbens influence cocaine-induced behaviours Nature Neurosci. 8, 413–414 (2005). Parkinson’s disease models. Nature 445, 643–647 as well as the rewarding impact of electrical Demonstration that self-administration of cocaine (2007). stimulation in the medial forebrain bundle. potentiates excitatory synaptic transmission in the 139. Day, M. et al. Selective elimination of glutamatergic 123. Zhang, X. F. , H u , X . T. & W h i t e, F. J. Whole-cell bed nucleus of the stria terminalis, but that passive synapses on striatopallidal neurons in Parkinson plasticity in cocaine withdrawal: reduced sodium administration of cocaine or food does not. disease models. Nature Neurosci. 9, 251–259 (2006). currents in nucleus accumbens neurons. J. Neurosci. 131. Grueter, B. A. et al. Extracellular-signal regulated 140. Lee, K. W. et al. Cocaine-induced dendritic spine 18, 488–498 (1998). kinase 1-dependent metabotropic glutamate receptor formation in D1 and D2 dopamine receptor-containing 124. Hu, X. T., Basu, S. & White, F. J. Repeated cocaine 5-induced long-term depression in the bed nucleus of medium spiny neurons in nucleus accumbens. Proc. administration suppresses HVA-Ca2+ potentials and the stria terminalis is disrupted by cocaine Natl Acad. Sci. USA 103, 3399–4304 (2006). enhances activity of K+ channels in rat nucleus administration. J. Neurosci. 26, 3210–3219 (2006). References 138–140 use BAC transgenic mice to accumbens neurons. J. Neurophysiol. 92, 132. Sigurdsson, T., Doyere, V., Cain, C. K. & LeDoux, J. E. demonstrate cell specific modifications of medium 15971–15977 (2004). Long-term potentiation in the amygdala: a cellular spiny neuron synapses in the dorsal and ventral 125. Dong, Y. et al. CREB modulates excitability of nucleus mechanism of fear learning and memory. striatum following in vivo manipulations. accumbens neurons. Nature Neurosci. 9, 475–477 Neuropharmacology 52, 215–227 (2007). 141. Zhang, F. et al. Multimodal fast optical (2006). 133. Childress, A. R. et al. Limbic activation during cue- interrogation of neural circuitry. Nature 446, 126. Tsankova, N., Renthal, W., Kumar, A. & Nestler, E. J. induced cocaine craving. Am. J. Psychiatry 156, 633–639 (2007). Epigenetic regulation in psychiatric disorders. Nature 11–18 (1999). 142. Zhang, F., Aravanis, A. M, Adamantidis, A., Rev. Neurosci. 8, 355–367 (2007). 134. Fu, Y. et al. Long-term potentiation (LTP) in the central de Lecea, L. & Deisseroth, K. Circuit-breakers: optical 127. Delfs, J. M., Zhu, Y., Druhan, J. P. & Aston-Jones, G. amygdala (CeA) is enhanced after prolonged technologies for probing neural signals and systems. Noradrenaline in the ventral forebrain is critical for withdrawal from chronic cocaine and requires CRF1 Nature Rev. Neurosci. 8, 577–581 (2007). opiate withdrawal-induced aversion. Nature 403, receptors. J. Neurophysiol. 97, 937–941 (2007). 143. Schultz W. Multiple dopamine functions at different 430–434 (2000). 135. Richter, R. M. & Weiss, F. In vivo CRF release in rat time courses. Annu. Rev. Neurosci. 30, 259–288 128. Walker, J. R., Ahmed, S. H., Gracy, K. N. & Koob, G. F. amygdala is increased during cocaine withdrawal in (2007). Microinjections of an opiate receptor antagonist into self-administering rats. Synapse 32, 254–261 the bed nucleus of the stria terminalis suppress heroin (1999). self-administration in dependent rats. Brain Res. 854, 136. Pollandt, S. et al. Cocaine withdrawal enhances long- DATABASES 85–92 (2000). term potentiation induced by corticotropin-releasing UniProtKB: http://ca.expasy.org/sprot 129. Weitlauf, C., Egli, R. E., Grueter, B. A. & Winder, D. G. factor at central amygdala glutamatergic synapses via BDNF | c-FOS | ERK1 | Orexin A | Orexin receptor 1 | PSD-95 High-frequency stimulation induces ethanol-sensitive CRF, NMDA receptors and PKA. Eur. J. Neurosci. 24, ALL LINKS ARE ACTIVE IN THE ONLINE PDF long-term potentiation at glutamatergic synapses in 1733–1743 (2006).